Hi,
I'm currently trying to use Maxent for roadkill analysis, but one of my colleagues raised a question that using spatially correlated datasets in maximum entropy models leads to biased results and the dataset should be tested and correlated occurrences removed.
I have no clue how to test roadkill point dataset for spatial autocorrelation to get rid of correlated points. What's the easiest and the most common way to determine which points need to be removed to avoid the bias?
Do you think we can dramatically reduce spatial autocorrelation (SAC) issue by selecting 'Random Subsampling' and creating 'Bias file' in MaxEnt settings? As you know, Maxent provides an option to use 'Random Subsampling' and also allows the creation of a Bias file that accounts for the distribution of the presence data. Please share your thoughts on this.
If you know / have any information / articles / website for me to look over, please advise on this. Your direct advise to solve this issue is greatly appreciated.
Best,
Bryan
The use of one of the common spatial autocorrelation metrics, Moran's I (ArcGIS) is well explained in Dubŏyk et al 2010 at RG (for GLM). The metric is applied to compare spatial models and reject those with a low spatial autocorrelation rather than a high one as predicted by the First Law of Geography.
Detection or removal of biased sample points can take several ways, including random sub-sampling (example attached: relative high number of points in autumn), if you have a large number of points (maybe >20 or >100 for Maxent).
If your independent variable, road kills is a proxy, e.g. for animal (gregarious or solitary?) distribution rather than road kill distribution, you may cull spatially and/or temporally close points (see attachment 1).
Spatial autocorrelation is also a scaling issue, dealt with implicitly in Dubŏyk paper. You may also run your SDM at alternative resolutions to see whether the model is stable/robust (attachment 1; done for attachment 2, but not included)
Article Fine resolution distribution modelling of endemics in Majell...
Article Where the bears roam in Majella National Park, Italy
Hi there
To avoid spatial autocorrelation of your occurrences points just filter the ones that fit in the same pixel. The pixel size will depend on your dataset layers. For example, if your pixel is 1km2 then you need to have only one occurrence point per 1km2.
hi Hoehun
you can watch videos on the following YouTube channel. I hope this channel helps you.
https://www.youtube.com/channel/UC-vVdGmKrGNveRDp3vskTOQ
Hi Hoehun
If you cab share with meyour method in select one point per grid using ArcGIS, or reducing spatial data in general.
Thanks in advance
Dear Hein Van Gils
Thank you so much for your helpful advice. Could you please share with us the complete citation of the work by Dubŏyk and colleagues?
Thank you in advance!
Sorry, I did misspell the name of the first author at the time. However, I did answer a similar question some years in this thread. Maybe you can trace this earlier answer. If not, let me know and I will try once more to provide a solid answer.
- Badges
- Science topic